Soft robotic manipulators with proprioception

Abstract

Agriculture and horticulture depend heavily on human labor to perform tasks that are often dirty, hazardous, and highly repetitive. One reason for the lack of automation of these tasks is the absence of suitable robotic handling equipment. Rigid robotic manipulators are typically incapable of performing dexterous manipulation tasks such as harvesting apples as they lack the ability to adapt to objects of various shapes and sizes. Such robotic manipulators need a large number of sensors and actuators to overcome these challenges, making them overly complex and not very robust. Therefore, the development of robotic manipulators for dexterous manipulation tasks has begun to focus on morphological computation, in which at least some aspects of the control are outsourced to the body of the robot. Taking inspiration from grasping mechanisms in natural systems, the field of soft robotics attempts to address this problem by constructing robots from soft materials. Although soft robotics may be the key to realizing automation of dexterous manipulation tasks, the current commercially available soft robotic grippers are only capable of performing simple pick-and-place tasks with open-loop control. This limited capability is in large part due to a lack of techniques to endow these manipulators with a sense of self-movement and body position, known as proprioception. Proprioception is a simple problem for conventional robots with rigid members and discrete joints, as the body position can be easily reconstructed using the information from encoders in the robots’ joints. However, it is a highly challenging problem for soft robots with virtually infinite degrees of freedom and above all, no suitable off-the-shelf sensors…